Abstract: The difficulty of grid-enabled application development is one of the obstacles to employ grid as infrastructure of scientific computing. Although there are many projects, especially grid workflow projects, trying to make grid-enabled application development easier and quicker, the process-based application description model adopted by most of the gridenabled applications still bothers scientists who are usually beginner in programming. In this paper, a data dependency based application description model was proposed, which describes a grid-enabled application via dependency among data items in the application, and much close to the thinking pattern of scientists. Based on this model, a scientific computing oriented grid application description language (SCO-GADL)was proposed, and an engine for supporting its execution was designed and implemented. The extensible plug-in architecture adopted by SCO-GADL and its engine makes them support different grid and non-grid platforms easily. At last, a test shows that grid-enabled scientific applications can be developed easily in SCO-GADI，and executed correctly on the application engine.

Key words: Scientific computing, Grid workflow, Description language, Data dependency